Normally if you decide to use some insertion point, different than centroid, you will be interested in getting the more accurate solicitations for your members, so you won't check the mark at "do not transform frame stiffness from offsets from centroid" because you would lose the accuracy.
Certainly the insertion point influences solicitations, imagine the steel part at some ordinary composite beam; placing everything at centroid wouldn't acknowledge the stiffness of the concrete at the compression head, nor that would help to some proper portrait of the stresses in the steel part. Hence insertion points must be used whenever we require bigger accuracy; we use them to get our models closer to the actual -yet always notional- structure in behaviour.
I have introduced above the word notional mainly to remember that our skeletal structures are but safe simplifications of what structurally we have come to consider relevant of our buildings. It is clear, for example, that structures with attached masonry will have a very much different stress status than at our calculations; normally great part of the flexural action will get diminished whereas our purportedly nonstructural material becomes stressed and our beams and columns behave far more axially. These "different than projected" behaviours are present anywhere as unrecognized structural behaviour, and only smart and sharp structural understanding will place them at its primary status (instead of the noninteresting to which normally -if devised- they are normally adjudicated), so if we want our structures to behave in some specific manner and our models recognize the design intent behaviour, we better use the tools at hand, one of which is the insertion point feature.
